hydraulics branch official file copyalthough flow patterns could be observed only at la•+...
TRANSCRIPT
HYDRAULICS BRANCH OFFICIAL FILE COPY
1 ` SYMPOSIUM 1970
STOCKHOLki i
' I ~ i
FREQUE14CY AND AMPLITUDE OF PRESSURE SURGES GENERATED
. BY SWIRLING FLOW
I H. T. Falvev U. S. Bureau of Denver, Colorado
I Heal, Hydraulic Research Section Reclamation USA
J. J. Cassidy University of Columbia, Missouri Professor of Civil Engineering Missouri USA
SYNOPSIS This paper reports on the investigation of swirling flow through tubes. Fre-
quencies and amplitudes of pressure surges produced by the swirling flow were
measured and found to be essentially independent of viscous effects for Reynolds
numbers larger. than 1x105. Dim;nsionless frequency and pressure par_.rrr2t2rs as
well as the onset of surging were correlated with the parameter SO!pQ2 Where sl and
Q are respectively the fluxes of angular momentum and volume through the tube, D
is the tube diameter and p is the fluid density. Relative length and shape of the
j tube were also found to be important.
The results of the study were used to analyze a particular draft tube for
potential surging. Using performance data obtained from the model of the turbine
and draft tube, a icyloil of sur-ye-free operation was outlined on the efficiency
I hill for the unit. Comparison was made between measured poorer swings and the
predicted pressure fluctuations in the draft tube.
RESUME Cette communication a trait a une recherche sur un courant tourbillonnaire
jpassant dans des tubes. Les frequences et les amplitudes des surpressions creees
j par le courant tourbillonnaire ont kt mesurees et trouvees essentiellement i -
indepandantes des effets de viscosite pour des nombres de Reynolds superieurs e
j 1x105. Des parametres sans dimension lies a la frequence et a la pression et
1'etablissement de la surpressions ont Le relie's au parametee QD/pQ2 ou ,i et Q
sont respectivement les flux du moment angulaire et du volume dans le tube, D le
diame-tre du tube et p la densite du fluide. On a trouve que la longueur relative
et la forme du tube sont egalement importantes.
On a utilise les resultats de cette etude pour analyser l'etablissement de la
surpressions dans une conduite d ecoulement particuliere. En utilisant les donnees
de fonctionnement obtenues a partier d'un modem de turbine et de conduite
d-ecoulement, on a ,rouve un domaine de marche mans surpressions sur le diagramme
de rendement de 1'unite. On a compare les resultats des mesures des amplitudes de
puissance et ceux des calculs des fluctuations de pression dans la conduite
i d'ecoulement.
INTRODUCTION
Draft-tube surging has been recognized as a problem in the operation of
hydro-power plants for several decades [1]. History shows that the attack on
draft-tube surging has been one of alleviation rather than prevention. New units
are placed in operation and if objectionable surging occurs within the operating
range various types of remedial action are attempted or, if possible, operation
is not permitted in the critical range. However, because of automation of the
generating systems, plants are frequently required to operate over a wide range
of turbine gate openings. Remedial measures including air admission to the draft
tube or the placement of fins in the draft-tube throat are only occasionally
successful. If it were possible to predict the possibility of surging prior to
installation and operation of a unit, considerable advantage could be realized.
It appears to be generally accepted that draft-tube surging arises as a
result of rotation remaining in the water as it leaves the turbine runner and
enters the draft tube [2]. Moreover, the phenomenon is complicated by many fac-
tors including vaporization of the water, geometry of the turbine and draft tube,
and the dynamic characteristics of the penstock and electrical network. Swirling
flow has been investigated analytically and experimentally by several people in
the field of fluid mechanics [3,4,5,6]. These studies have shown that when a
fluid flows in a swirling fashion (axial flow with superimposed rotation) the
J
resulting flow pattern depends upon the relative amount of rotation in the
J fluid.
If the flux of angular momentum entering the tube is sufficiently large as com-
pared to the flux of linear momentum, a reversal in flow direction occurs along
the centerline of the tube [3]. For large Reynolds numbers the flow becomes
unsteady forming a helical vortex with the reversed flow occurring along the
spiral core of the vortex [7]. This phenomenon is known as "vortex breakdown" in
the fluid mechanics literature and as draft-tube surging in the hydraulic-machin-
ery literature [8].
This paper describes a study initiated in order to investigate the character-
istics of swirling flow in tubes and to correlate these characteristics with the
occurrence and nature of draft-tube surging.
THE Rncrr cTiinv
Because of the excessive number of variables which could be involved it
was not only desirable but necessary to simplify the experimental model as much
as possible. Air was used as the working fluid in order to eliminate the possible
effects of a two-phase flow. No turbine runner was installed in order to simplify
geometry. Instead the air entered the tube (Figure 1) radially through wicket-
gate type vanes. It was possible to accurately determine the amount of swirl
imparted to the flow as it entered the tube. The angle between the vanes and a
radial line could be set at any angle between 0° and 82°. At 00 the vanes were
2 g 1
its ig i
aft
)e, iny-
i
id.
in in-
ter-
ie
L
Flow -tea Wicket Gate or Inlet Vane
Smoke Injecter Here _
Flow _
D
-enter Line
Stilling Chamber
FIGURE 1 EXPERIMENTAL APPARATUS SHOWING STRAIGHT TUBE
or
a
FIGURE 2 DEFINITION SKETCH OF INLET FLOW TO TUBE OR RUNNER ible
Iify
1 3
Ue
radial and no rotation was applied to the flow (Figure 2). Overall configuration
of the apparatus is shown in Figure 1 and is described in more detail by Cassidy
[7]. Clear plastic was used in construction to facilitate flow visualization
with smoke. Pressures at points on the tube wall were measured using pressure
cells and a root-mean-square meter. Frequencies of the unsteady pressure were
determined using an oscilloscope with a retentive screen.
Flow through three types of tubes was studied:
1) circular tubes of uniform diameter; 2) a model of a draft tube used in
Fontenelle Dam; 3) an expanding cone having the same length-to-area relation-
ship as the draft tube.
ANALYSIS
It was assumed that, for a particular draft-tube shape, the root-mean-square
amplitude AP and frequency f of the unsteady pressure surge are both functions of
the fluid density p and viscosity v, draft-tube diameter D and length L, discharge
Q, and flux of angular momentum Q. Standard techniques of dimensional analysis
yield the following functional relationships
D4AP = ~ /1Z D L IR ) (1) ~2 s n2 — D
7c,93 3 = ~2(p D ) ~ ) (2)
wherelR is the Reynolds number 4Q/nDv.
The angular momentum flux Q enterinq the tube was computed as (see Figure 2)
11 to OR V. sin a (3)
in which Q = SBNVo.. The variables S, a, Vo, and R are defined in Figure (2);
N is the number of vanes around the inlet; and B is the dimension of the vanes
perpendicular to the plane of Figure 2. The momentum parameter contained in
Equations (1) and (2) is then computed as
rLD _ DR sin a (4) pQ _ BNS
and is seen to be a function of geometry only for the particular inlet conditions
of this study. In all experiments the momentum parameter nD/pQZwas calculated
according to Equation (4).,
~s 4 S 1
PA
V%
a.) Steady Swirling Flow, no b.) Stagnation Point at the c.) Fully Developed Surging, surging, D=3.46 inches, Centerline With Surging SID /pQ2=0.454, D=6.13 inches, L/D=7.15, QD/pQ2=0.150 downstream D=3J6 inches, L/D=3.26.
L/D=7.15, SID/pQ =0.330
Figure 3. SWIRLING FLOW IN STRAIGHT TUBES
RESULTS
Flow patterns at low velocities were readily made visible with smoke. With
the inlet vanes in the radial position no rotation was imparted to the fluid and
streamlines in the test section were essentially straight and parrallel to the
tube centerline. With the vanes inclined slightly from the radial position
(discharge held constant) streamlines in the test Section became steady spirals.
That condition is shown in Figure 3a. Further closure of the vanes eventually
produced vortex breakdown. Figure 3b shows the stagnation point occurring near
the right quarter point of the tube with a helical vortex occurring downstream
from the stagnation point. Still further closure of the vanes moved the stagnation
point to the upstream limit of the tube. Figure 3c shows the helical vortex
filling the tube. A comparison of this helical vortex pattern.with those observed
below a turbine runner [2,8] establishes the similarity.
Although flow patterns could be observed only at la•+ velocities (below 5 feet
per second) unsteady pressures could be measured only at higher velocities. To
insure that similar flow patterns occurred at these higher velocities, a hot-
wire anemometer was used to measure air velocities near the tube wall and at the
tube centerline. These measurements showed that surging began at the same gate
setting (same QD/pQ2) regardless of discharge.
In the experiments pressure amplitudes and frequencies occurring after surg-
0 ing commenced were measured near the open end of the straight tubes, upstream
from the elbow of the draft tube, and near the entrance of the straight cone. The
onset of surging (vortex breakdown) was correlated with the value of QD/pQ2. It
was also found to be a function of L/D and shape. For the elbow draft-tube and
the straight cone the critical value of RD/pQ2 was 0.4. Above this value surging
occurred.
The dimensionless frequency and pressure parameters of Equations (1) and (2),
calculated from the experimental measurements, are shown in Figures (4) and (5).
These parameters were found to be independent of viscous effects for Reynolds
numbers greater than 1 x 105 and the results for smaller Reynolds numbers are not
shown here.
In order to obtain reproducible results, it was necessary to measure the
root-mean-square values of the pressure fluctuations and it is these values which
were used in the preparation of Figure 4. Straight tubes are seen to have
pressure characteristics which are strongly influenced by relative length. The
pressure characteristics of the elbow draft-tube and straight cone are striking
since the pressure parameters exhibit a maximum value as nD/pQ2 is made larger.
Frequencies of the unsteady pressure were quite regular and, thus, relatively
easy to measure. The regularity indicates that the helical vortex precesses about
the tube centerline at a reasonably regular rate. Figure (5) shows that the
divergent tubes significantly reduce the frequency of surging. However, the bend
in the draft tube has an insignificant effect upon frequency.
6 g 1
g~—
+ + +
2.0 Symbols Defined ++ + ;,1x
in Figure 5 - tCylindersl +4- x
+ x
+ °°° x Cone + .6p D4
o -1.29[I:1.108(P 2-1.41)] t x PQ2
1.0 . o
Ap D4 x °
• x • x x x
x
Q2 P x '• x
''-' XxElbow Draft Tube xA 4
xxxx P0=1.111:1.563 42 C
;;7 xx
0 0 1.0 12 D 2.0 3.0
PQ2 Fig. 4-PRESSURE PARAMETERS FOR SURGING
x x +
2.0 x +
x+ + I x + -
X + j
f D3 x +
x ~
Q ++
o° 1 °°
x °°O
1.0
•po as
D L/ D .521 ft. (15.9 cm) 4.68 Elbow Draft Tube
ebb ° .503ft.(15.3cm) 4.84Truncafed Cone 4-.511 ft (15.6 cm) 3.26 Circular Cylinder x .511 ft, (15.6cm) 1.63 Circular Cylinder
0 All Tubes Have Sharp edged Entrcnce5
0 1.0 a D 2.0 3.0 i
Fig. 5- FREQUENCY PARAMETERS FOR SURGING
7 E 1
ANALYSIS OF DRAFT-TUBE SURGING IN A TURBINE
• Occurrence and severity of draft-tube surging in a given hydraulic turbine
and associated draft tube can be predicted using the experimental results of this
study. To do this it is necessary to be able to compute the parameter PD/pQ2 of
the flow as it leaves the turbine runner and enters the draft tube.
Analysis of Turbine
Horsepower P imparted to the turbine runner by water flowing through it is
given by the classic expression
P rr -n z)15 (5) 60 In '
where n is the rotational speed of the turbine in revolutions per minute, and
Q2) is the rate of change in angular momentum occurring in the flow as it
passes through the runner. Subscripts 1 and 2 refer to the entrance and exit
sides of the runner respectively. Multiplying both sides of Equation (5) by
550 D/npQ2 and using Equation (4) yields
f12 D DRsina _550 P„ D ( 6 ) — P92 BN5 2V2-9,' P DZ
where ~, Pll, Qll, and D2 are respectively the specific speed, power, discharge
and runner diameter and are defined in Figure 6 and Q2D/pQ2 is the momentum para-
meter associated with the flow leaving the turbine runner and entering the draft
tube. The first term on the right of Equation (6) is the momentum parameter
associated with the flow leaving the wicket gates and entering the turbine and
can be evaluated if the wicket-gate geometry is known. The second term on the
right of Equation (6) can be evaluated if the performance characteristics and
9s geometry of the turbine are known..
Specific Application
Equation (6) was used to analyze the hydraulic turbine at Fontenelle Dam
[9]. The efficiency hill for the turbine, as determined from model tests, is
shown in Figure 6. Any point on this efficiency hill yields a particular value
Of Q2D/pQ2 and, thus, areas of potential surging--where E12D/pQ2 exceeds 0.4--can
be determined. Figure 7 shows the regions in which surging can be expected and
also shows a line representing operating conditions for which flow leaving the
turbine and entering the draft tube has zero swirl. To the left of this line flow
in the draft tube swirls opposite to the turbine rotation while to the right
swirl is with the runner. The point of maximum efficiency falls on the line of
zero swirl for this unit although this is not apparently true of all units.
Figure (4) predicts that for the elbow draft tube a maximum pressure ampli-
tude occurs at Q2D/pQ2 = 1.30. The line on which surging would produce this
maximum pressure amplitude is also shown on Figure 7. A parabolic curve was fit-
ted to the pressure data for the elbow draft tube. This parabolic expression
(shown in Figure 4)was used in conjunction with Equatim(6)tocalculate relativevalues
8 $ 1
O D2
i
090
80 17= 0/6Efficie cy 76
0 /
50 a 30 2 10 Runaway
~0 Speed P
P11=D2 HBO: e
N S = ~g ¢ P 2 =63 0 ~— n
0 1.0 7rD2 n 2.0
¢ 601. 2gH FIGURE 6 EFFICIENCY HILL FOR FONTENELLE TURBINE
2 .0
'17 900
F /
~c
c •may ~r • y i
o I.0 Zero Swirl—, g~
_ - Runaway v / Speed
\Maximum Surge Amplitude
01 0 1.0
De n 2.0
¢ 60'V29
FIGURE 7 PREDICTED SURGING FOR FONTENELLE TURBINE
9 $
(AP/pgH where H is net head on the unit) of pressure amplitudes for the Fontenelle
unit. The results of this computation are shown on Figure 8 along with relative
power swings actually measured at the plant. Excellent agreement is shown for
the wicket gate opening at which maximum surging occurred. Numerical agreement
of the % power swing and % pressure fluctuation was not expected because the
dynamics of the mechanical and electrical system are necessarily involved in
determining the power swing arising as a result of the pressure surging which is
the driving force. The lack of agreement occurring at each side of Figure 8 may
also be due to the use of Equation (4). Analysis of other units indicated that
this expression may not give an entirely accurate estimate of the momentum para-
meter for all gate openings usually because the exact geometry of the wicket gate
assembly is not accurately known.
CONCLUSIONS
A particular hydraulic turbine can be analyzed for possible surging within
the desired range of operation by the method presented herein. Dimensionless
frequencies and root-mean-square amplitudes of unsteady pressures measured herein
are directly convertible to prototype conditions by similitude considerations.
However, they can be expected to predict prototype conditions only for the "hard-
,r core" vortex condition. For low-tailwater conditions, for which the vortex core
becomes unwatered, a two-phase flow occurs and pressure surges are usually some-
what milder.
ACKNOWLEDGEMENTS
Mr. U. J. Palde of the Hydraulics Branch, U.S. Bureau of Reclamation, took
much of the data and carefully refined the experimental methods and equipment
used.
PPPrPrNrrcz
1.) Rheingans, W.J., "Power Swings in Hydroelectric Power Plants," Transactions,
American Society of Mechanical Engineers, Vol. 62, 1940.
2.) Hosoi, Y., "Experimental Investigations of Pressure Surge in Draft Tubes of
Francis Water Turbines," Hitachi Review, Volume 14, Number 12, 1965.
3.) Harvey, J.K., "Some Observation of the Breakdown Phenomenon," Journal of
Fluid Mechanics, Vol. 14, 1962.
4.) Chanaud, R.C., "Observations of Oscillatory Motion in Certain Swirling Flows,'
Journal of Fluid Mechanics, Vol. 21, 1965.
5.) Benjamin, T.B., "Theory of the Vortex Breakdown Phenomenon," Journal of
Fluid Mechanics, Vol. 14, 1962.
6.) Squire, H.B., "Analysis of the Vortex Breakdown Phenomenon, Part I," Aero
Department, Imperial College, Report Number 102, 1960.
10 B I
Fontenelle
relative
)wn for
ireement
the
!d in
which is
!re 8 may
ed that
um para-
cket gate
within
iless
M herein
Lions.
ie "hard-
:ex core
y some-
, took
ment
Rated Output=11.9 Mega-watts ~r~Predicted % Pressure Fluctuation
02.0 4 Net Head= 100 Feet
z
z ,Measured % Power ._0 1.5 3 Swing-
Peak to Peak
o + \ +
\
ti 1.0 ° 2 i
o S \
0.5 I / \
H o /
4 0 0 +
0 20 40 60
% Wicket Gate Opening
FIGURE 6 PREDICTED AND MEASURED SURGING EFFECTS
actions,
:bes of
I of
ig Flows,'
of
Aero
Contribution by 1. Raabe on Paper E1
Since the publication of Meldau:"Drallstrdmung im Drehhohlraum" in 1935
we know that steady potential flow in a straight tube with an axial
component and a circulation as Dr Falvey and Prof. Cassidy describe; is
usually connected with a dead water core.The ratio of core diameter
to internal tube diameter is controlled by the law of minimum of
kinetic flow energy content in the tube. This law is a special case
of the law of minimum resistance. Bamme-A and Kl&L&eY)s have enriched
1950 our knowledge of straight tubular potential flow with circulation
in so far as they stated and observed a back flow in the core for
those cases in which the pressure on the external boundary of the core
drops below the back pressure. In a longer tube with pressure losses
in axial direction this return of the flow in the core may happen at
variable stations on the axis.The position of this depends on circu-
lation and thrnughfloW.
to reality the dead water core is also set in~ otation by boundary
friction.Therefore it, can be taken as a vortex.Slight irregularities
in the axial flow with enlarged discharge displace the vortex center
from the axis. These irregularities belong strongly to the inlet
conditions in the tube.
Hereby the displaced vortex core element induces at other vortex core
elements a radial displacement. The displaced element does the same
with a following element, by this forming a helical vortex. This vohtex
rotates in circumferential direction in the sense of circulation of
the surrounding flow, also due to induced velocity. The tangential
induced velocity at any core element and its approximation to the wall
due to suction effect of surrounding flow enlarges the radial deviation
of the vortex core so that it touches the tube wall at last.This is
seen clearly by the 'experimehts of Dr Falvey and Prof. Cassidy.
They stated that viscosity the observed Reynolds number-off 105
is negligible. Therefore-rotational speed of thie--vortex and tie pressure
pulsations in connection herewith can only be controlled_by inertia—___.__
r
forces of rotating and axial-flow masses. Dimensionless pressure
amplitude and its frequency is therefore a function of dimensionless
ratio of kinetic energy of rotational and axial flow as formulae 1
and 2 show.
The dependence of the ratio Z/D and Reynolds number describes the
dependence of the flow returning point in the axis as function of
pressure drop as mentioned above.
The curiousity not known in real turbines seems to be that the f ormatic
of the helicoidal whirl begins at large distances from the inlet of
the tube. It is in contrast to the swirl of a turbine, beginning mostly
behind the runner, at the hub or in the runner itself.
The use of the model formula in the shell diagramm of a turbine
therefore rises some doubts:
1/ Flow in a turbine runner, especially a Francis runner of high
specific speed, can only be described by one circulation before and
behind the runner after thoroughful measurements of flow velocity at
these stations.
2/ Due to fig 4 pressure pulsation will be higher on a cylinder. This
in contrast to the experience that an internal coaxial cylindrical
tube in the draft tube quietens the pressure pulsations remarkably.
3/ Fig 7 shows surging regions but not the magnitude of pressure
pulsation. The limits of these regions are gained by equations 6
which uses the shell diagramm fig 6. This diagramm doesn't distinguish
a helical whirl region and a pulsating straight whirl region.
Especially the pulsations of a central whirl at higher discharges
with circulation against the speed causes the stronger shocks.
4/ From fig 5 follows for a runner with given D,N,B that
- - - -- — w-here W,,* gu e opening angle. is t e
For Francis turbines one can estimate
cos p(o — constant
By this
f H Observing
fn u
we have
f/n - constant
Concluding I can state that the air tests of the authors have a
preliminary value in predicting qualitatively the ep-region but not the exact strength of Qp. On behalf of a constant frequency/
speed ratio the paper may give some approximation to the prototype.
The resting guide apparatus with its wicket gates and its sharp
inlet edges in the tube doesn't coincide with the runner as the
real boundary condition.
SYMrU51UM 1970, STOCKHOLM
H. T. Falvey, J. J. Cassidy, "Frequency and Amplitude of Pressure Surges Generated by Swirling Flow"
t
DISCUSSION
by Rex A. Elder, Director,.TVA Engineering Laboratory, USA and Svein Vigander, Research Engineer, TVA Engineering Laboratory, USA
.The authors have presented model data in dimensionless form
from which the frequency and amplitude of pressure pulsations in
draft tubes of hydraulic turbines can be predicted. This type of
information is needed both to better understand the generation of pe-
riodicity in the swirling flow and to enable prediction of the fre-
quency and severity of disturbances to be expected in new turbine designs.
We would like to add a few comments based on the studies
made by TVA's Engineering Laboratory on the vibrations of a 36-MW
fixed-bladed turbine unit. Mention of these vibration studies were
made earlier, at the 1968 Lausanne Symposium and at the 13th Congress
in Kyoto.
On the enclosed Figure 1, we show the results of measure-
ments of pressure pulsations made at the draft tube man door below
the turbine runner. The pressure pulsations were measured with a
flush-mounted, diaphragm-type, electronic pressure transducer. The
data has been reduced to the dimensionless coordinates shown on the
author's Figure 4 and Figure 5. The applicable parts of these fig-
ures are also shown on the enclosed Figure 1. The prototype data
scatter right around the curves drawn through the model results for
both frequency and amplitude of pulsations. This agreement, we feel,
is quite remarkable, particularily in view of the fact that geometric
similarity between the model and the prototype in this case was lim-
ited to whatever is implied in the general phrase "Elbow Draft Tube".
In light of these results, it would be interesting to see data from
Fontenelle Dam and other dams plotted in a similar manner. We hope
that the addition of this information may make engineers even more
confident in the use of the author's results.
LL; 2.0 Lv , O U Q SNL
LLJ CURVE THROUGH 1.5 FALVEY AND CASSIDY'S
1— MODEL DATA s WHEELER UNIT 11
p GAGE P4 ON DRAFT
Q I.0 TUBE MANDOOR D
N p I SYMBOL TEST DATE O pp O ID p I O E2 9/67
N_ 0'S _
0 0 O E 8/67
C 1 8/67
0 0 0 0.5 1.0 1.5 2.5 2.5
I 100 70 50 40 30 25
PERCENT GATE OPENING
Mr. S. GASAuc;i
Doctor of Math Sciences Technical Director Energy Division
ALSTHOM - NEYRPIC
Grenoble - France
Remarks on report E1 of Mr. H.T. FALVEY and J.I. CASSIDY "Frequency and amplitude of pressure surges generated by
swirling flow"
The paper of Mr. Falvey and Mr. Cassidy comprises an interest-ing attempt to explain pressure surge phenomena which are pro-duced in the draft tubes of hydraulic turbines ; this explana-tion gives an account of the general aspect of their experimen-tal results.
On the characteristic hill curve of a Francis turbine, can indeed be seen a surge-free zone corresponding to slight swirling flow at the runner outlet. Outside this zone, swirl-ing flow appears rotating in the same direction as the runner or in the opposite direction,depending on whether the operatin,_ point happens to be above or below the surge-free zone.
Starting from the hill curve and knowing 521, it would be pos-sible to calculates 2, but it.is in fact very difficult to calculatenj from the geometry of the distributor. Moreover, Ao defined from the output corresponds to an overall value. There exist operating points where the water, leaving the run-ner, turns in one direction for one operating zone and in the other for another. When swirling flow appears at the runner outlet, what value must be chosen for n2 to define S22 D/pQ2:
On either side of the "calm" area, having an absolute compara-ble value n2 D/pQ2, the contoursof the axial velocities differ widely ; the stability conditions should not therefore be the same ; furthermore the distribution of the axial velocities in the model without a runner used by the authors is very diffe-rent from those of a model with a runner. It is worth pointing out that the vortices do not take the same form on each side of the swirl free zone : they are roughly axial for the large Q11 and helical for the small Q11. It seems therefore difficult to characterise the instability of a flow so complex by a sin-gle parameter SZD/pQ2.
The attached figure shows the distribution of axial velocities; at the draft tube inlet for various operating conditions. For points above and below the "calm" zone, the maximum axial velocities for this machine occur at the centre and near the draft tube wall respectively. For a givenSftD/pq2, the stabili-ty conditions for these two types of flow should be different.
It therefore seems that tests on _a model equipped with a run-ner are necessary to define the pressure surges and their transposition onto the prototype machine. The frequency of these pressure surges is of as much if not more interest than their amplitude. In fact, the problematical pressure fluctua-tions do not always have well defined frequencies. For partial loads, more or less periodic fluctuations can be noted ; the frequencies of the maximum amplitudes depend on the machine's speed of rotation and are little influenced by cavitation. For high loads, the frequencies, less well defined, are a functior of T. Their problematical character would require the use of the energy spectral densities and root mean square values. On this subject, the authors compare the peak-to-peak amplitude variations of the power fluctuations to the root mean square values calculated from the pressure surges. In fact, the root mean square values should be compared, allowing for the trans-fer function linking the variations in output to the variatiot in pressure in the draft tube, considered as the only importai disturbances. Readings taken on industrial machines do seem tc indicate that the spectral densities relating to shaft torque variations are difficult to link to those concerning draft tui pressure fluctuations.
Finally, it must be stressed that it is the whole range of frequencies of several tens of Hertz and under which is of practical interest and not the lowest frequencies of about 1 Hertz. Knowledge of this frequency range also requires test! on a complete turbine model.
r
a,) ofri
- - - ~• ~ X ~~Ilaljirilgr
~i.,~l 1 ~Nlti
- ~% i`-
~ 13
66*76
84.
i I 4 r
• j j i ~ O 11 t
r — Sens rc
DISCHARGE DISTRIBOTIC
,AND FLOW ROTAT ION BE LOW THE RUNNER
r
DISCUSSION TO PAPER El BY CASSIDY AND FALVEY:
Authors' Reply by John J. Cassidy and Henry T. Falvey
It is both frustrating and rewarding to receive a wealth of
discussion to one's paper. We have demonstrated a basic approach
to the analysis of an old problem. Because the problem is indeed
complex and has received attention from many investigators over
at least four decades, it was natural that both favorable and un-
favorable comments should arise. The authors' contention is simply
that draft-tube surging is a fluid-flow phenomenon and as such must
by analyzed in terms of fluid-flow parameters. The hvdraulic tur-
bine, as a machine, changes not only the energy content of the water
as it passes through the runner but the momentum content and dis-'
tribution as well. We will attempt to answer each discusser's com-
ments in the light of our contention.
The prototype measurements plotted in accordance with our pa-
rameters and presented by Elder and Vigander were quite encouraging.
However, in their Fig. 1 (Draft-tube pressure pulsations) the curve
denoted as "curve through Falvey and Cassidy's model data" should
not pass through the origin of the plot of f Dyp versus rL D/I.Q? The critical value of .C1D1 QZ as measured by the authors was 0.4
below which surging did not occur. A drawing showing the geometry
of the Fontennelle draft is included in this reply. Unquestionably
the shape of the draft tube has an effect upon both the occurrence
and the magnitude of pressure surges. This effect of shape is one
which should receive considerably more attention,
NOHABfs measurements as presented by Mr. Holmen provide an ex-
tremely gratifying support and simultaneously illustrate the impor-
tance of the plant 6 or cavitation coefficient. The effect of 6
was alluded to in t^e author's paper but could not be supported with
data. Mr. HoL„en's results show that as 6 decreases the amplitude
of the pressure surge decreases. It is the author's contention
that as 6 is reduced the flow becomes two-phase to an increasing
degree. Because of the rotation in the draft all air and vapor mi-
grate to the tube- center: . Appa-ren-tly this -e-f~-ect stabilizes the vortex. This 6 effect is in keeping with the field observation that
A, ,
rging is alleviated when tail-water elevation is decreased or when
L, is injected into the draft tube. The authors deliberately chose
r as the medium for our study in order to eliminate C as one of
a important parameters. However, detailed study of the effect of
6 is certainly desirable and is a logical second step in the inves-
tigation of draft-tube surging.
Dr. Casacci makes several interesting points. In a given hydro-
electric plant it is possible for resonance to occur in the penstock,
the electrical network, the governor system, or in a combination of
all. However, the draft-tube surge frequency is the driving fre-
quency and its use is mandatory in a study of possible resonance
conditions. For the Fontenelle-dam surge tanks prevented penstock
resonance and the plant output of approximately 10 megawatts is
negligible as compared to the connecting system. Resonance was not
expected and, thus, it is not surprising that- the maximum power
swing coincided with the gate opening producing the maximum ampli-
tude of pressure surge.
The calculation of SLZD0 QZ seems to raise difficulty but not in the way Dr. Casacci questions. Equation (6) is a momentum equation
and is, thus, affected only by forces which produce a torque. Re-
gardless of what the velocity distribution is at the draft-tube en-
trance ilzQ~po" will be correct if S11 D~PQ'' and turbine characteristics are correct. Equation (4) gives a reasonable value for 111 D/14L in all cases we have studied.
The effect of velocity distribution on the critical value of
tLZp,P is certainly a valid question. It was also of considerable
concern in our study. In our complete study (Journal of Fluid
Mechanics (1970), vol. 41, part 4, pp. 727-730) we deliberately
varied-the inlet geometry so as to vary the inlet-velocity distri-
bution. No apparent effect on the critical value of R,D~PQZ was
noted. We conclude, therefore, that there is a unique flow pattern
for a given value of n2D/PQZ even though the development of the flow pattern may differ. - " -
Dr. Casacci's statement that dynamic characteristics of shaft
torque characteristics care. difficult to link to the frequencies of
the draft-tube surges is in contradiction to other studies. One
thing we have often noted is that pressure taps in prototype draft
tubes frequently are located where eddies shed from upstream geometry
are measured rather than the pressure surge of interest. Also, the
dynamic effect of the connecting electrical system frequently plays
an important part in determining the generator motion and hence the
shaft torque.
Professor Raabe questions essentially the validity of our con-
tention that "vortex breakdowns' and draft-tube surging are one and
the same. Our observations as seen in Fig. 3 show the flow pattern
in each to be the same. Simultaneous measurements of pressure and
velocity show that the unsteady pressure is due to an unsteady
velocity. Observations of flow patterns at low velocities and pres-
sures at high velocities show that surging occurs at the same value
of-Q2. 0/10 regardless of the magnitude of Q (with some allowances
for Reynolds number effect). Prototype measurements and independent
measurements in other laboratories support our contention. Therefore,
we rest our case.
Placing a coaxial cylinder in a draft tube alleviates surging
because it effectively decreases -UD11oQz since the effective diameter
D is decreased.
Finally, some observations should be made relative to further
study of draft-tube surging. Fixed-blade turbines will always impart
angular momentum to the flow entering the draft tube at some settings.
It is logical that some changes in runners can be made to improve on
surging characteristics. This, however, will probably reach the
point of diminishing returns quite rapidly. It would appear that
changes in draft-tube design offer the greatest possibility for
improvement. In this light, the effect of divergence of the draft
tube is currently under study. However, even more imaginative
studies should be pursued. Presently, draft-tube designs are satis-
factory only when the turbine is operating at maximum efficiency.
However, operation is most frequent at conditions other than maximum
' efficiency for which flow is frequently upstream in one half of the
draft tube. For high-head plants, surging might well be avoided
entirely if no-draft tube--were provided and efficiency would- be
changed very little percentage-wise. In short, it would seem that
less attention could be paid to the peak efficiency and more to
other operationai cnaracteristics.
D~ 10.0'
Aff
4 M
~ 3 Pier Gate
Ki Nose slot
2
A~ 3- / / Half Cone Angle =
L /D3
2 3 4 5
- -1.03 D3
D3~
Figure/ Fontene/% Draft Tube